How Adult Content Moderation Works

Behind the word «moderation» sits a conveyor of several stages. Understanding how it is built explains most refusals and suspensions that look arbitrary from outside.

Tier 1. Hash matching

Every uploaded file is compared against databases of known prohibited material by digital fingerprint. The comparison survives recompression and cropping. This is the fastest and harshest filter: a match means immediate blocking and a report to the relevant organisations.

Tier 2. Automatic classification

A model assesses the content: type, indications of minority, violence, prohibited categories. It works quickly, errs in both directions, and is always tuned with a margin towards refusal — the cost of letting prohibited material through is incomparable with the cost of a false rejection.

Tier 3. A human

Contested cases and complaints are handled by people. This is the conveyor's narrowest point: slow, expensive and hard on the moderators themselves. Which is exactly why platforms try to resolve as much as possible at the first two tiers.

Documentation

Serious platforms require confirmation of the age of everyone in a shoot and consent to publication — before upload, not after a complaint. Material without documentation does not pass regardless of content.

Synthetic content

The newest and fastest-changing part of the rules. The general trend is to require labelling of generated material and to prohibit synthetic images of real people without their consent. Wordings differ; the direction does not.

How to reduce the risk of refusal

Keep documentation in order, label synthetic material yourself, and do not argue with automation at first instance: an appeal to a human is almost always more productive than a repeat upload.

This material is intended for adults over 18 and is informational in nature.